Uhm. I was bored, so thought I would roam around the web. Found reference to this pic from nvnews :
http://www.trinisica.com/img/galler...ge=stilllife-translucent_21_2.jpg&w=713&h=451
Info page for the pic :
http://www.trinisica.com/sub_view_imgpage.asp?loc=3&img=stilllife-translucent
Basically it's photorealistic image done with only basic 3D Studio Max scanline rendering... Catch here is, that it seems that nVidia will use this pic as to demo what nv30 will be able to render in _real-time_. After all, it's simple scene with straightforward high quality texturing.
Now, i know there have been some speculation about the nv30, but I still would like to hear any ideas about the what to expect. That is, if we have 120 million transistors, what you can put there ?
Like how much four floating point vertex shaders would eat from the silicon budget ? It's obvious that nv30 will implement dx9 pipeline, which is fully floating point, right ?. But how much more it will be ?
Or is it just a straightforward 4 vertex shaders , 4 pixel pipes with total of 16 textures / pass, better HSR, support for internal 64-bit floating point rendering and boring 128 bit bus...
In many ways nVidia has already build the hype like many times before , "Biggest thing in 3d gfx for last ten years" etc. It would be rather lame if nv30 turn out to be something like suped up gf4. Even with high quality mpg decode.
Oh, and do you think that nv30 will have conditional stament + jump in shaders ? That is, will the pixel shaders computers be fully Turing-complete with floating point support ? This is not dx9 specs, but is in OGL 2.0.
http://www.trinisica.com/img/galler...ge=stilllife-translucent_21_2.jpg&w=713&h=451
Info page for the pic :
http://www.trinisica.com/sub_view_imgpage.asp?loc=3&img=stilllife-translucent
Basically it's photorealistic image done with only basic 3D Studio Max scanline rendering... Catch here is, that it seems that nVidia will use this pic as to demo what nv30 will be able to render in _real-time_. After all, it's simple scene with straightforward high quality texturing.
Now, i know there have been some speculation about the nv30, but I still would like to hear any ideas about the what to expect. That is, if we have 120 million transistors, what you can put there ?
Like how much four floating point vertex shaders would eat from the silicon budget ? It's obvious that nv30 will implement dx9 pipeline, which is fully floating point, right ?. But how much more it will be ?
Or is it just a straightforward 4 vertex shaders , 4 pixel pipes with total of 16 textures / pass, better HSR, support for internal 64-bit floating point rendering and boring 128 bit bus...
In many ways nVidia has already build the hype like many times before , "Biggest thing in 3d gfx for last ten years" etc. It would be rather lame if nv30 turn out to be something like suped up gf4. Even with high quality mpg decode.
Oh, and do you think that nv30 will have conditional stament + jump in shaders ? That is, will the pixel shaders computers be fully Turing-complete with floating point support ? This is not dx9 specs, but is in OGL 2.0.